Document Server@UHasselt >
Research publications >
Please use this identifier to cite or link to this item:
|Title: ||Réformulation des question dans l'interrogation des bases des donneés bibliographiques : Système Lexiquest|
|Authors: ||Pomian, J.|
|Issue Date: ||1988|
|Citation: ||Egghe, L. & Rousseau, R. (Ed.) Informetrics 87/88, Belgium : Diepenbeek, p. 197- 208|
|Abstract: ||The purpose of the LEXIQUEST project is to introduce the use of artificial intelligence techniques in information retrieval. LEXIQUEST handles relations between indexing key-words. It helps the user to find out a meaningful formulation of his query, which is stated in free terms.
Most database interfaces provide the user with catalographic tools (lists of titles, authors, key words, and so on), but are unable to detect the user's preoccupation. Some systems, for example DIALECT or EXPRIM, use an interpretation of the user's request to reformulate his query. This is also the purpose of LEXIQUEST, but our system differs in that the procedures and the rules are independent from the field on which the program is run, since they are directly generated by specialised programs. It is therefore possible to run the LEXIQUEST system on any key-word indexed database, without modifying the system.
LEXIQUEST is designed to guide any query made on a key-word indexed database. The system interactively refines and reformulates the initial query, using the relations it finds between key-words. After some iterations, the final formulation is validated by the user, and the relevant documents are retrieved.
The links between key-words are weighted by the "E" coefficient, in order to detect each word's semantic environment. The value of this "E" coefficient is :
1) E = C i . C.
This "Equivalence" coefficient measures the mutual inclusion between two words i and j, by using their frequencies Ci and Ci, and their co-occurrence Ci.
The system maces an intensive use of the transitivity of the "E" coefficient. Links weighted with this coefficient are interpreted as rules, and are recorded as such in the set of rules used by the system. The LEXIQUEST system furthermore uses a saturation metarule in order to limit the size of the network attached to each word.
The LEXIQUEST system is interesting in two points. First of all, there is no use in creating manually the association network for each database, since it is created automatically. The system can therefore
be adapted to any database. Second, there is no need to know the indexing vocabulary in advance, since LEXIQUEST can handle the reformulation of the query.|
|Type: ||Proceedings Paper|
|Appears in Collections: ||Research publications|
Files in This Item:
|N/A||395.67 kB||Adobe PDF|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.